He, he,
yes that's how water cooling is working, we know that but...
How loud is the pump? How big that system needs to be - PC level or car level for typical class A amp we are using? These are the specific that will determine how usable that is. If I have to have a loud pump that is equal to me as a forced air. In that case neither one will fit. But if someone was able to use quiet PC system that is easily available with obviously low prices - several manufactures are offering it, than that might be a way to go. I have no idea what PC cooling systems are designed to handle, but I doubt they will be able to cool even 2 pairs per side of typical Mosfets biased to 0.6V on the source resistor.
yes that's how water cooling is working, we know that but...
How loud is the pump? How big that system needs to be - PC level or car level for typical class A amp we are using? These are the specific that will determine how usable that is. If I have to have a loud pump that is equal to me as a forced air. In that case neither one will fit. But if someone was able to use quiet PC system that is easily available with obviously low prices - several manufactures are offering it, than that might be a way to go. I have no idea what PC cooling systems are designed to handle, but I doubt they will be able to cool even 2 pairs per side of typical Mosfets biased to 0.6V on the source resistor.
When the packaging allows for 350degC operation, we can just let the JFET boil water and release the steam to outdoor.
Then the amp will be consuming deionized water other than electricity.
A 1000W consumption amp will convert 1.59kg of water into steam every hour.
Seems like a lot.
So maybe recycle the water will be better.
Anyway it will be pretty easy to convert steam to water, compared to cooling warm water.
Plus, no pump needed.
But wait... Isn't this gas cycle exactly same as the way heat pipes work?
So we came back to the origin, just use a finned cooler with heat-pipe build in will do the job well.
Then the amp will be consuming deionized water other than electricity.
A 1000W consumption amp will convert 1.59kg of water into steam every hour.
Seems like a lot.
So maybe recycle the water will be better.
Anyway it will be pretty easy to convert steam to water, compared to cooling warm water.
Plus, no pump needed.
But wait... Isn't this gas cycle exactly same as the way heat pipes work?
So we came back to the origin, just use a finned cooler with heat-pipe build in will do the job well.
Last edited:
Well it's not nonesense. But, it may be difficult to take advantage of these principles better than what has already been done. For instance the nucleate boiling principle works fine. Using water however dictates your temp regulation at 100C (212F). I would be happier with my packages a little cooler than that. You just use a fluid with lower bopiling point 
Still, kinda messy though.
Heatpipe, cpu coolers are interesting. Most all use forced air cooling though. To reduce the air noise you would need to slow down the fan and derate. Most of these units are made for processors from 80-150W. It's easy to use a TO247 film resistor with known characteristics and a power supply to test these things...
I think a self powered, auto circulating, sealed, water cooling idea has the most promiss. The PC guys already do that too. 
Still, kinda messy though.Heatpipe, cpu coolers are interesting. Most all use forced air cooling though. To reduce the air noise you would need to slow down the fan and derate. Most of these units are made for processors from 80-150W. It's easy to use a TO247 film resistor with known characteristics and a power supply to test these things...
I think a self powered, auto circulating, sealed, water cooling idea has the most promiss. The PC guys already do that too.
Yes, i wanted to point ou that the fanless thermal resistance of the HR-01 is quite poor, probably unable to dissipate 50W, and raises dramatically with a very little fan.
I should have known that that's where you were going. And i'm sorry but that's not my cup of tea.
I have tried an undervolted Nexus fan, such as the ones in SPCRW reference tests, which should be 18-19dB and i HEAR IT at 1.9m of distance wit closed case.
So any fan in hifi equipment is a total turnoff for me.
Watercooling unfortunately also need a fan. You can look into TEC cooling, with external units that may do without fans.
Take a look at FLG's answer, that's what I'm talking about. No pump needed.
As for the 100C figure, why would that be a problem regarding the Semisouth devices?
As I understand it, the only reason to keep them as low as 100C is the plastic of the package. The die should be good for like 700C as I recall.
...and no TeaBag, this is not nonsense, but quite advanced development, as we for once actually can use the nuclear boiling principal, using plain water, for cooling a device, as the device can deal with the higher temp.
Think about it, and you will find that all of a sudden, quite small boxes becomes an option.
Magura
I think there is a few points that need to be brought out:
The 50W passive figure is for CPU, which you need to keep the die at <85degC or so to keep the CPU running. With SiC JFET that can do 175degC, you can easily double if not triple that figure. When we have packaging capable of 400+degC, we will be able to [COLOR="#SeaGreen"]6X[/COLOR] this CPU figure.
HR-01 and such, when installed in typical ATX case, will be in the orientation that the wind should blowing to the back of the case. To make passive cooling work better, we should rotate the mounting by 90deg and be in a position that the wind should be blowing vertically up or down. Then hot air will rise with less resistance and exit faster, making passive cooling better.
And another point is stack effect. The reason why passive cooling is so poor is that hot air does not gather enough momentum to rise fast.
Stack effect says that the speed of passive air flow is proportional to the square root of the stack height. With HR-01, the native stack height when pointing up, is only 5cm. if we add a 50cm chimneys on top of the HR-01, we can get more than [COLOR="#SeaGreen"]3x the air flow[/COLOR]. And that make a significant difference, with no additional noise.
HR-01 and such, when installed in typical ATX case, will be in the orientation that the wind should blowing to the back of the case. To make passive cooling work better, we should rotate the mounting by 90deg and be in a position that the wind should be blowing vertically up or down. Then hot air will rise with less resistance and exit faster, making passive cooling better.
And another point is stack effect. The reason why passive cooling is so poor is that hot air does not gather enough momentum to rise fast.
Stack effect says that the speed of passive air flow is proportional to the square root of the stack height. With HR-01, the native stack height when pointing up, is only 5cm. if we add a 50cm chimneys on top of the HR-01, we can get more than [COLOR="#SeaGreen"]3x the air flow[/COLOR]. And that make a significant difference, with no additional noise.
Multiply the 3 points above, it will not be difficult to get 8x (or 20x with 400degC packaging) the passive dissipation capability as in the unmodified passive CPU cooler usage. How about 400W (or 1000W with 400degC packaging) per 12x12x5cm/600g heatsink passively cooled?
Last edited:
as we for once actually can use the nuclear boiling principal, using plain water, for cooling a device, as the device can deal with the higher temp.
Magura
The boiling just move the heat from device to steam, which we still need to get rid of later.
And this is no different from how a heatsink with heatpipe works.
1. liquid absorb heat and evaporate
2. gas move to the far ends of the aluminum heatsink.
3. gas condense to liquid, and the heat is removed to the air.
4. the liquid move back to the hot spot by gravity or capillary action.
So having device capable of doing water boiling is really nothing special, unless you just throw away that steam, replenish the deionized water, and forget about using aluminum heatsink at all. But that will take quite a bit of water every hour, as I calculated earlier.2. gas move to the far ends of the aluminum heatsink.
3. gas condense to liquid, and the heat is removed to the air.
4. the liquid move back to the hot spot by gravity or capillary action.
I don't exactly have "the answer" however it is possible to maintain a prescribed case/die temp by emersing the power device in an inert (high resistivity) fluid. There is no "free lunch", you have to disperse of the heat somewhere somehow? But you can build a small tight electronic circuit with close tracking temp properties, and efficiently send the heat problem some where away from the electronics. The nucleate boiling has a natural regulation property about the fluids boiling point. Something even an active heatsink system dreams about.
My experience is with high power digital ASICs that were being "burned In" at 170C. This was thought to be 17 times life acceleration. They were bare die approximately 1cm sq. suspended on .002" capton tape. They varied from 7.5-23watts dissapation ea.. We did 48 hours of burn in at 170-175C replicating 1 month of typical operation. The die had a diode specifically designed for temperature monitoring of the Vf vs. temp. One thing I feel guilty about never working out would be a noise I beleived to be the actuall boiling transmitted to the diode that indicated die temp. This would be a killer to us if we wanted to emerse bare die, The signal would develope thermal distortion effects manifesting themselves as boiling noise
I think a package would efffectively filter those effects? There is a Thj situation with the pakage but Magura is probably right. 100C is cool Yea, well, you know us North Americans and being to late... But, I recall we could probaly get away with 125-145C on the polymer plastic crap.
Yes, this is mostly correct. The fluid I experimented and used was classified as a solvent. The "thermal cunductivety" was approximately equal to water but the resistivity was very high probably like highly spec'ed "clean" "DI" or otherwise water? It was "inert" we could put an entire pcb operating into this fluid. The fluid was specifically selected for it's boiling point to maintain our process control. BTW, I have worked in several design facilities where 170C kicks in the overtemp circuits affecting a shutdown. Mother Nature's clean pure water is somehow a more sane, elegant, solution here. I understand the infamous mil spec (would that be the US military?) dosen't like their die temp over 100 or 105C (someone else must know this garbage?) or something. But that is normal Si not the magical SiC things we are about to experiment with... I would be interested in a comment from the all-knowing one, N.P., regarding this phenomina??? Have you pushed them a little yet? Is there a benefit to linearity or sound?...this is not nonsense,,, using plain water, for cooling a device, as the device can deal with the higher temp.
Think about it, and you will find that all of a sudden, quite small boxes becomes an option.
Magura
I would gladly donate selected portions of my brain, it's experience and it's time to someone who wishes to experiment in this area further...
I would also suggest refreshing your memories with the infamous Grey Rollins Water Cooled Heat Sink Thread and any associated hits on such ideas. When he see's my spelling he might squeeze out from the woodwork anyhow???
Last edited:
Your points are correct, but in the SPC review here, it must be noted also:
- that vertical mounting does NOT help as much as thought.
- we dont know which was the max cpu temp beyond 85°, maybe around 100°, maybe more.
In conclusion, i think a practical test to measure the cooling power and the transistor Tj may be useful before investing in such cooling.
Thank you Telstar for pointing out these.
I also notice that I made another mistake.
I assumed thermal dissipation power increase linearly with temperature delta between the heatsink and the ambient.
This is only correct when the speed of air is fixed.
In reality speed of air will rise with the square root of the hot/ambient air temperature difference. This increase in air speed will cool down the heatsink, but also decrease the hot/ambient air temp difference which cause this increase in air speed in the first place. The equivalent point will be that for N times increase in the hot/ambient air temperature difference, the heat dissipation for the same heatsink will increase by N^1.5 times.
This means that, although the manufacturer's rated maximum power only increase linearly with the maximum junction temperature, but for passive cooling with heatsink using stack effect, the power that we can safely dissipate, for the same heatsink and devices having the same RTHjc, will be proportional to (maximum die temperature - temperature delta between the die and the heasink - ambient air temperature)^1.5.
This show that, the increase die temperature of SiC device make them extremely suitable for passive cooling with stack effect.
Let's see if I will have an opportunity to do a test with HR-01 Plus and SJEP120R100 (the RTHjc of SJEP170R550 is too big for good test). I'm looking for each HR-01 Plus to dissipate about 150W passively with multiple insulated SJEP120R100.
I also notice that I made another mistake.
I assumed thermal dissipation power increase linearly with temperature delta between the heatsink and the ambient.
This is only correct when the speed of air is fixed.
In reality speed of air will rise with the square root of the hot/ambient air temperature difference. This increase in air speed will cool down the heatsink, but also decrease the hot/ambient air temp difference which cause this increase in air speed in the first place. The equivalent point will be that for N times increase in the hot/ambient air temperature difference, the heat dissipation for the same heatsink will increase by N^1.5 times.
This means that, although the manufacturer's rated maximum power only increase linearly with the maximum junction temperature, but for passive cooling with heatsink using stack effect, the power that we can safely dissipate, for the same heatsink and devices having the same RTHjc, will be proportional to (maximum die temperature - temperature delta between the die and the heasink - ambient air temperature)^1.5.
This show that, the increase die temperature of SiC device make them extremely suitable for passive cooling with stack effect.
Let's see if I will have an opportunity to do a test with HR-01 Plus and SJEP120R100 (the RTHjc of SJEP170R550 is too big for good test). I'm looking for each HR-01 Plus to dissipate about 150W passively with multiple insulated SJEP120R100.
Last edited:
Zhoufang, what amps these jFets are used in and that are readily available for the DIY? I believe you somewhere mentioned X configuration. What amp, Aleph X?